Social media crises, such as the Samsung Galaxy Note 7 recall, unfold rapidly and generate intense public reactions. To respond effectively, organizations require tools that go beyond sentiment detection to capture underlying emotional patterns. However, existing sentiment classification approaches often prioritize accuracy while struggling to adapt to domain-specific crisis contexts. This study proposed a Multitask Learning (MTL) framework for jointly classifying sentiment and emotion in Reddit-based crisis communication. Transformer-based models, including BERTweet and DistilRoBERTa, were fine-tuned under both Single-Task Learning (STL) and MTL configurations using benchmark datasets and evaluated on a curated Reddit corpus from the Note 7 crisis. To ensure robustness, performance was assessed using Macro-F1 as the primary evaluation metric, with standard deviation reported across five random seeds to capture performance robustness and stability. The results show that STL models consistently outperform MTL on general-domain benchmarks for emotion classification, whereas MTL offers modest improvements for sentiment classification in crisis-specific data. Emotion classification remains particularly challenging, with most models consistently achieving lower Macro-F1 scores. Overall, the findings highlight both the potential and limitations of MTL for crisis analysis and provide a foundation for developing more reliable and transparent artificial intelligence tools to support effective crisis communication.